Imagine it's a post-Covid world. I know it's hard to imagine, but let's start with the feeling of going into a movie theater. You're in a movie theater with all of your best friends watching a new scary movie on the big screen. This is the first movie you haven't watched on your home TV in a long time. When you get out of the theater, it’s dark and you can’t imagine walking home because—let’s admit it—you’re scared.
You whip out your phone to get a ride from Uber. It’s late at night on a Friday and everyone is out and about enjoying their lives post-quarantine, so the price is high—higher than you remember the price being on a pre-Covid Friday night when you went to go see “Once Upon a Time in Hollywood.” You ask your driver why the prices are so high and they say “The demand is higher. Everyone is out more now after quarantine. Also, that doesn't mean you shouldn't tip me.” You two share an awkward silence.
What they really mean is that Uber is using “dynamic pricing.”
What is Dynamic Pricing?
Dynamic pricing is a way to base prices on current market conditions. Uber does it. Airbnb does it. Airlines do it too. Basically, anything that has rising and falling prices is probably using a dynamic pricing method.
Dynamic Price Methods
- Cost-based: Based on business costs to keep profit margins consistent.
- Competitor-based: Set to keep prices competitive.
- Demand-based: Set based on growing or shrinking demand.
Do You Need a Complex ML to Set Up Dynamic Pricing?
Good question. You can use historical pricing data and plug them into machine learning algorithms to predict how much a customer is willing to pay at certain times. For companies that have a lot of daily transactions between customers, machine learning is invaluable.
The more data a company has, the more accurate dynamic price points will be. Additionally using machine learning provides more flexibility than a rule-based system due to the ability to change its output with a changing environment. As you feed your ML system current data, it can tell you real-time predictions and price services accordingly.
Use Cases of Machine Learning and Dynamic Pricing
Using Obviously AI, a tool made for non-technical business users, you can setup this dynamic pricing without writing technical code OR having any background in ML.
If you’re coding an algorithm yourself or taking a more technical approach, I recommend this post.
Uncover Hidden Relationships Between Data Points
Crossing over into the telecommunications industry, we plugged in AT&T customer data into Obviously AI’s platform to see if we could get some insights. Let’s look at the relationship between churn and value of monthly charges. This helps figure out the factors related to churn and how much customers are willing to pay with certain services.
Obviously AI automatically built, tested, and deployed an algorithm tailored specifically for this dataset. For this instance, the algorithm is called the Gradient Boosting Regressor to figure out that the monthly charges were directly proportional to churn. We can also figure out those who had bank transfer payment method, are female, and had a two-year contract had a higher possibility of churning.
Segment Customers Based on their Characteristics
From the data, we can also conclude this is the average amount a customer is willing to pay before they churn. From these and other characteristics related to churn like age, internet service, if they stream movies or not, etc., can help us build personas of customers.
Predict Customer Purchasing Behavior for Price Optimization
Another thing you can learn from this data is customer behavior based on altered services. Businesses can visualize which service features customers are willing to pay higher for.
For this use, we plugged in Airbnb data to determine what characteristics customers will pay a higher price for. With our platform, you can conclude neighborhood type, availability, room type, and minimum nights affect the price.
Customers were willing to pay the highest when they rent the entire home on Roosevelt Island in Manhattan with the average number of reviews for the home being 16.70.
With this data, you can pick the optimized price based on the characteristics of the services you’re providing.
If you want to learn more about creative data predictions, check out this post we published a few weeks back.
The Possibilities With ML and Dynamic Pricing Doesn’t Stop Here.
There are just so many things you can do and methods you can use to get the most optimized price for your service using ML. We wanted to show you some examples to spark your creativity in searching for business solutions with predictive analytics.
If you have any more ideas or want to discuss this post, follow us on Twitter.